Manual Before Automated — Process Hygiene as the Foundation of AI Workflows
The Insight
Before you automate any process with AI, you must first work it out manually, practice it manually, and only then move it into an automation. Automating a poorly understood or poorly designed process doesn’t speed things up — it locks in the mess at scale.
This was stated directly by Don Back in the October 30 chat: “Work out the process behaviour, practice the behaviour manually, then and only then move it into an automation.” He added the cautionary note: “I’ve had to unwind automated messes that did not build off of clear processes.”
Why It Matters
The temptation with AI tools is to automate early — to hand a task off before you understand it well enough to evaluate the output. This creates two compounding problems:
- You can’t debug what you’ve never done. If you haven’t done a task manually, you have no reference point for evaluating whether the AI output is correct, good, or even in the right direction.
- Bad process, amplified. AI automation doesn’t fix a broken workflow — it executes it faster, at higher volume, with consistent errors.
The manual phase serves a specific function: it generates the process knowledge that the automation needs. Without it, you’re not delegating — you’re outsourcing your ignorance.
The Principle Applied
This maps directly to a hierarchy of AI readiness:
- Understand the task — can you articulate the goal, the inputs, the quality criteria?
- Do it yourself — execute the process manually at least once end-to-end
- Refine through repetition — identify the friction points, judgment calls, and edge cases
- Then automate — now the automation has a clear spec, and you can evaluate its output
The “AI as team member” framing (also from this session — “building my team one $20 agent at a time”) only works when you know what you’d brief a human team member to do. Process clarity is the prerequisite.
Relevance for Coaches and Knowledge Entrepreneurs
This principle is especially important for coaches building AI-assisted client delivery or content systems. The common pattern of failure:
- Coach discovers a new AI tool
- Attempts to automate client intake, content creation, or follow-up without a clear manual process
- Gets inconsistent output, confused clients, or automation drift
- Concludes “AI doesn’t work for my practice”
The real cause: the underlying process was never clearly designed. The AI revealed the gap, it didn’t create it.